Consensus-Based Weights
Consensus-based weights, also known as liquid alpha, make validator-miner bond smoothing depend on how validator weights align with consensus. The mechanism sits inside the Yuma Consensus bond path between validator-weight signals and validator-side outcomes (Consensus-based Weights, Yuma Consensus).
The core idea is alignment-sensitive smoothing. Stronger consensus alignment can make the related bond response more direct, while weaker alignment keeps the response more conservative.
That places the mechanism on the bond-response side of Yuma Consensus. Validator weights provide evaluation signals; consensus-based weights adjust how validator-miner bonds respond after those signals are compared with consensus (Consensus-based Weights, Glossary: Validator-Miner Bonds).
Liquid Alpha
Liquid alpha is another name for consensus-based weights. The name highlights that the alpha term in bond smoothing can vary rather than staying fixed across every validator-miner pair (Consensus-based Weights, Exponential Moving Averages).
Liquid alpha and consensus-based weights name the same mechanism. They do not name separate reward types or separate miner-scoring systems.
Here, alpha is the EMA smoothing parameter, not a subnet alpha token. That keeps the mechanism tied to bond smoothing rather than token ownership (Exponential Moving Averages, Consensus-based Weights).
This terminology matters because alpha-token vocabulary is common elsewhere in Bittensor. Consensus-based weights use alpha for smoothing responsiveness, not pool-held alpha, alpha outstanding, or stake.
Bond Smoothing
Bittensor uses exponential moving averages to smooth validator-miner bonds, so bond state does not change entirely from one evaluation step to the next. Consensus-based weights make the smoothing factor responsive to consensus alignment (Exponential Moving Averages, Consensus-based Weights).
That places the mechanism after validator weights are compared with consensus. It does not replace weight input, clipping, rank calculation, or Yuma Consensus itself.
The smoothing role is narrow. Consensus-based weights affect how validator-miner bond history responds, while the surrounding Yuma Consensus path still defines how weights, ranks, and emissions are interpreted.
That makes the mechanism a response-shaping step. It changes how quickly bond history reacts after consensus comparison, while validator weights and the consensus process still provide the input side.
The important movement is in the bond history, not in the original validator-weight submission. A validator still submits weights; consensus-based weights change how the bond path reacts around alignment (Consensus-based Weights, Yuma Consensus).
Validator Dividends
Consensus-based weights affect the bond path that feeds validator dividends. Yuma Consensus routes validator emissions through bonds to miners weighted by miner emission results, and liquid alpha adjusts how quickly those bonds respond to consensus-aligned evaluation (Yuma Consensus: validator emissions, Consensus-based Weights).
The result is still validator-side consensus behavior, not a new miner-ranking rule.
That distinction keeps dividend language from becoming a broader performance label. The mechanism changes bond responsiveness; it does not directly rewrite the task being evaluated by a subnet.
The dividend connection is therefore downstream. Consensus-based weights adjust the bond state that feeds validator-side emissions, while dividends name the validator-side result (Yuma Consensus: validator emissions, Glossary: Validator-Miner Bonds).
Weight Copying
Consensus-based weights are often discussed near the weight-copying problem because copied or stale weights can diverge from consensus. The mechanism can make bond response more sensitive to whether validator weights remain aligned (The Weight Copying Problem, Consensus-based Weights).
This does not make liquid alpha an accusation model. Weight copying names an evaluation-quality problem; consensus-based weights name a bond-path response to consensus alignment.
Development Stage Context
The Introduction to Bittensor describes subnet development as moving from localnet to testnet and then mainnet. For consensus-based weights, that sequence changes how readers should interpret bond-smoothing and liquid-alpha examples.
In localnet, consensus-based weight mechanics can be tested in an isolated environment. Localnet bond history does not represent production validator-miner relationships.
On testnet, liquid-alpha bond smoothing can be exercised in a shared non-production network. Testnet consensus alignment outcomes are separate from mainnet subnet state.
On mainnet, consensus-based weights apply as a live Yuma Consensus bond-smoothing step on production subnets. Observed bond response depends on the selected subnet’s validator weights and consensus context (Consensus-based Weights).
The Bittensor Networks reference separates mainnet, testnet, and localnet. A liquid-alpha example from one environment should not be read as representing production bond behavior in another environment.
Relationship to Yuma Consensus
Consensus-Based Weights and Yuma Consensus describe related parts of Bittensor’s incentive system. Yuma Consensus is the on-chain process that aggregates validator weight signals within a subnet into miner incentives and validator dividends, applying consensus clipping, bonding, and emission calculation (Yuma Consensus).
For readers, consensus-Based Weights names a specific part of that incentive picture, while Yuma Consensus names the consensus process that turns validator weights into the resulting incentives and dividends.
Reader Boundary
Consensus-based weights are a Yuma Consensus bond-smoothing concept. They are not a separate subnet token, a new miner task, or a complete explanation of validator rewards (Consensus-based Weights, Glossary: Validator-Miner Bonds).
The mechanism only becomes meaningful with the subnet, incentive mechanism, and consensus process that interpret validator weights (Understanding Incentive Mechanisms, Yuma Consensus).
Clipping Filters Weights Before Bond Smoothing Responds
Consensus-based weights are adjacent to clipping but not identical to it. Clipping limits outlier validator weights above the consensus benchmark before final rank calculations, while liquid alpha adjusts how bond history responds after weights are compared with consensus (Glossary: Consensus Score, Yuma Consensus).
Clipping is a filtering step and liquid alpha is a bond-smoothing step. Both sit inside the consensus path, but they answer different questions about validator-weight signals.
Commit Reveal and Liquid Alpha Address Different Copy Risks
Delayed visibility through Commit Reveal addresses timing around visible weights, while consensus-based weights address bond responsiveness around alignment (Commit Reveal, Consensus-based Weights).
The relationship is complementary. Commit Reveal reduces the value of copying fresh weights before they are revealed; consensus-based weights change how bonds respond when submitted weights align or fall out of alignment (Commit Reveal, The Weight Copying Problem).